spiritfather/Melinoe-Qwen3-8-27B-VL-i1-GGUF overview

---

base_model: bgg1996/Melinoe-Qwen3-8-27B-VL

language:

• en

library_name: transformers

license: apache-2.0

quantized_by: spiritfather

tags:

• text-generation
• creative-writing
• gguf
• imatrix

---

About

weighted/imatrix quants of https://huggingface.co/bgg1996/Melinoe-Qwen3-8-27B-VL

> Benchmarked on CaliperBench — a creative-writing benchmark scoring prose craft, roleplay and willingness rather than general intelligence. See this model's scores: caliperbench.com/m/melinoe-qwen3.8-27b-vl.

> Static (non-imatrix) quants of this model are at spiritfather/Melinoe-Qwen3-8-27B-VL-GGUF.

> imatrix (importance matrix) computed over bartowski calibration_datav3. Weighted quants spend precision where it matters most — the low bit-rates (IQ1–IQ4) are meaningfully better than same-size static quants; at Q5/Q6 the difference is negligible.

> Vision: this is a VL model. llama.cpp keeps the vision tower OUT of the main GGUF, so multimodal use needs mmproj-Melinoe-Qwen3-8-27B-VL-BF16.gguf alongside your chosen quant (llama-server -m <quant>.gguf --mmproj mmproj-Melinoe-Qwen3-8-27B-VL-BF16.gguf). Text-only use needs just the quant.

<!-- provided-files -->

Usage

If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files.

Provided Quants

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

| Link | Type | Size/GB | Notes |

|:-----|:-----|--------:|:------|

| GGUF | Q2_K | 10.9 | |

| GGUF | Q3_K_S | 12.3 | |

| GGUF | IQ3_S | 12.6 | |

| GGUF | IQ3_M | 12.8 | |

| GGUF | Q3_K_M | 13.5 | IQ3_S probably better |

| GGUF | Q3_K_L | 14.6 | |

| GGUF | IQ4_XS | 15.3 | |

| GGUF | IQ4_NL | 16.0 | |

| GGUF | Q4_0 | 15.8 | |

| GGUF | Q4_K_S | 15.8 | optimal size/speed/quality |

| GGUF | Q4_K_M | 16.8 | fast, recommended |

| GGUF | Q4_1 | 17.3 | |

| GGUF | Q5_K_S | 19.0 | |

| GGUF | Q5_K_M | 19.5 | |

| GGUF | Q6_K | 22.4 | practically like static Q6_K |